Messages in this thread | | | Date | Wed, 16 Sep 2020 18:34:52 +0200 | From | osalvador@suse ... | Subject | Re: [PATCH v3 0/5] HWpoison: further fixes and cleanups |
| |
On 2020-09-16 18:30, osalvador@suse.de wrote: > On 2020-09-16 16:46, Aristeu Rozanski wrote: >> Hi Oscar, >> >> On Wed, Sep 16, 2020 at 04:09:30PM +0200, Oscar Salvador wrote: >>> On Wed, Sep 16, 2020 at 09:53:58AM -0400, Aristeu Rozanski wrote: >>> Can you try the other patch I posted in response to Naoya? >> >> Same thing: >> >> [ 369.195056] Soft offlining pfn 0x3fb5bf at process virtual address >> 0x7ffc84350000 >> [ 369.195073] page:000000002bb131e4 refcount:1 mapcount:0 >> mapping:0000000000000000 index:0x7ffc8435 pfn:0x3fb5bf >> [ 369.195080] anon flags: >> 0x3ffff80008000e(referenced|uptodate|dirty|swapbacked) >> [ 369.202131] raw: 003ffff80008000e 5deadbeef0000100 5deadbeef0000122 >> c000003fda1c7431 >> [ 369.202137] raw: 000000007ffc8435 0000000000000000 00000001ffffffff >> c000003fd63af000 >> [ 369.202141] page dumped because: page_handle_poison >> [ 369.202145] page->mem_cgroup:c000003fd63af000 >> [ 369.215055] page_handle_poison: hugepage_or_freepage failed�n >> [ 369.215057] __soft_offline_page: page_handle_poison -EBUSY >> [ 369.215068] page:000000002bb131e4 refcount:3 mapcount:0 >> mapping:00000000f6ca3f32 index:0x5c pfn:0x3fb5bf >> [ 369.215110] aops:xfs_address_space_operations [xfs] ino:49f9c5f >> dentry name:"messages" >> [ 369.215117] flags: 0x3ffff800002008(dirty|private) >> [ 369.215121] raw: 003ffff800002008 5deadbeef0000100 5deadbeef0000122 >> c000003fadd3daa8 >> [ 369.215127] raw: 000000000000005c c000003fd9497c20 00000003ffffffff >> c000003fd1143000 >> [ 369.215132] page dumped because: __soft_offline_page after migrate >> [ 369.215136] page->mem_cgroup:c000003fd1143000 > > > Ok, this is something different. > The race you saw previously is kinda normal as there is a race window > between spotting a freepage and taking it off the buddy freelists. > The retry patch should help there. > > The issue you are seeing right here is due to the call to > page_handle_poison in __soft_offline_page being wrong, as we pass > hugepage_or_freepage = true inconditionally, which is wrong. > > Should be:
Fat fingers, sorry:
Ok, this is something different. The race you saw previously is kinda normal as there is a race window between spotting a freepage and taking it off the buddy freelists. The retry patch should help there.
The issue you are seeing right here is due to the call to page_handle_poison in __soft_offline_page being wrong, as we pass hugepage_or_freepage = true inconditionally, which is wrong. I think it was caused during rebasing.
Should be:
@@ -1858,8 +1903,11 @@ static int __soft_offline_page(struct page *page) if (!ret) { bool release = !huge;
- if (!page_handle_poison(page, true, release)) + if (!page_handle_poison(page, huge, release)) { + pr_info("%s: page_handle_poison -EBUSY\n", __func__); + dump_page(page, "__soft_offline_page after migrate"); ret = -EBUSY; + }
Could you try that on top please?
I am away from my laptop now but I will be taking a look later today.
thanks
| |